Purchase this article with an account.
Ou Tan, Xinbo Zhang, Nils Loewen, Joel Schuman, David Greenfield, Rohit Varma, David Huang, ; The Effect of Image Quality on the Reliability of Nerve Fiber Layer Measurements with Fourier-Domain OCT. Invest. Ophthalmol. Vis. Sci. 2013;54(15):4820.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine if image quality affects the reliability of peripalliary retinal nerve fiber layer (NFL) thickness.
One hundred and nineteen normal and 136 glaucomatous eyes were scanned with optical nerve head (ONH) scan using RTVue Fourier-domain optical coherence tomography (OCT) system (Optovue, Fremont, CA). Each eye was scanned 3 times every visit. Normal subjects were scanned every 6 month and glaucoma subjects every 6 months. Scans from the earliest available 2 visits in the multicenter Advanced Imaging for Glaucoma study (www.AIGStudy.net) were used. Overall NFL thickness was obtained using RTVue 184.108.40.206 software. Marginal signal strength, defined as signal strength index (SSI) between 35 and 45, was detected using the automated RTVue software. Axial or transverse image cropping artifacts were detected using an automated software developed by coauthor Tan. The scans were classified as “weak-signal” (SSI=35-45), “cropped” (cropping artifact detected), or “clean” (SSI>45 and no image cropping). The reliability of NFL measurement was assessed by inter-subject variance in normal eyes, intra-visit repeatability, and inter-visit reproducibility, all evaluated by pooled standard deviation.
In the normal group, clean scans had significantly smaller inter-subject variance (p<0.001 and p=0.01, F-test) in overall NFL thickness than cropped and weak-signal scans (Table 1). Cropped and weak-signal scans had poorer reproducibility and repeatability than clean scans (p<0.001 for all comparisons by F-test, Table 2).
Excluding low quality scans (SSI<45, cropping artifacts) may improve the reliability of glaucoma diagnosis and monitoring by OCT NFL thickness measurements. Automated software to detect image cropping may be useful.
This PDF is available to Subscribers Only